1. . , . , 2014 . , . , -, 2014 . 1 22

2. 1 2 3 Differential evolution 4 No free lunch theorem . , . , -, 2014 . 2 22

3. F0 = arg max F p(FjX) + ; F; . . , . , -, 2014 . 3 22

4. dream team; ; . . , . , -, 2014 . 4 22

5. . ) ) "" ) ) , "" : . . , . , -, 2014 . 5 22

6. 1 2 "" 3 1 2 3 "" 4 "" . , . , -, 2014 . 6 22

7. - / - - - "" "" - "" - . , . , -, 2014 . 7 22

8. 1 2 "" 3 1 2 3 "" 4 "" . , . , -, 2014 . 8 22

9. . : [0; 1]n f0; 1gk . , . , -, 2014 . 9 22

10. - ! : , "" shuffle "" "" ( ) :) . , . , -, 2014 . 10 22

11. , . ! "" : n-point crossover; cutnsplice; . . , . , -, 2014 . 11 22

12. . , . , -, 2014 . 12 22

13. (genetic drift) ) . MCMC/ . . , . , -, 2014 . 13 22

14. vs - "" , - "" . : , . , , . , . . . , . , -, 2014 . 14 22

15. % "" "penalty" . , . , -, 2014 . 15 22

16. pros cons , . : ( :)); ; ; , . : ; ( ); . . , . , -, 2014 . 16 22

17. . , . , -, 2014 . 17 22

18. Differential Evolution arg max

19. 2Rn F(

20. ) 1 2 , x 2 P: 1 a; b; c 2 P ; 2 k U(1::n); 3 y = (yi ) i 1 r U((0; 1)) 2 yi = ai + F(bi ci ); i = kjr < C yi = xi 4 , . 3 . . , . , -, 2014 . 18 22

21. 2 . ? : : : , : Theorem (No free lunch theorem) . , . , -, 2014 . 19 22

22. NFL: dm = f(dxm (1); dym (1)); : : : ; (dxm (m); dym (m))g f : X ! Y F = YX p(f ) = 1F Theorem (David Wolpert and William G. Macready (1997)) a1 a2: P f p(dym jf ;m; a1) = P f p(dym P jf ;m; a2) f p(dym jf0;M;m; a1) = P f p(dym jf0;M;m; a2) . , . , -, 2014 . 20 22

23. NFL: :) , . , . , -, 2014 . 21 22

24. , : ln(x) : . , . , -, 2014 . 22 22